Fun fact: there used to be an Authy flatpak that just installed the snap inside
Oh, what the fuck!?
TBH I wouldn’t mind it that much. The whole point of flatpak is that the developer can do whatever demented satanic rituals they want inside of the sandbox, and it won’t contaminate the rest of the system.
2FAuth. On the web so you can check it anywhere you want and supports passkeys.
Ente Auth > Authy
Yep. I’m selfhosting it now. Works great but selfhosting isn’t straightforward yet, still the best Authy/Google/Microsoft Authenticator drop in replacement with sync.
I think I’ve landed on Flatpak as my favourite between Snap, Flatpak, and AppImage. AppImage, when it works, is nice though. Snaps are just kind of inconvenient (auto-updates are a no for me) and bloated and the things Canonical are doing as an organization put a bad taste in my mouth.
Auto-updates are a hell-no for me.
There was a perfectly good user interface for updates. Then Ubuntu decides “wait… What if we made updates compulsory and effectively random and skipped the UI. The user can do system updates whenever they want, because those don’t matter for security or something, but these apps must be updated whenever snap determines they must.”
Oh, snap!
I used to use flatpak for everything, but I just dont have the hard drive space to store duplicates of my graphics drivers.
I’ve had bad experiences with AppImages. For universal format they do a really poor job at that. And it’s a huge step back into Windows direction that you’ll have to manually download, update etc your shit. Makes managing a bunch of apps a pain.
Yeah, same here, that’s why I specified that they’re only nice when they work. Often they just don’t work, so Flatpak is better.
The thing with appimages is that they expect the developer to have full knowledge of what libraries need to be bundled with their app, which makes it difficult to make truly universal appimages. In flatpak you just select one of a set list of runtimes and add any additional dependencies on top of it. Flatpak also re-uses the files for each runtime in between the different apps that use it, which saves a lot of disk space.
But isn’t appimage the closest one to the app-system from Android? Since things could be really different on many clients an “app-container” is the best solution.
Why not containerise everything? You need libreoffice? No problem, here is a docker or podman container.
BTW. I like flatpak, too. It’s the most stable, but I never understand it’s mechanics. There is always another pack installed, freecode, gtk, qt whatever. Even if the system has already the correct gtk version, nope, the dev decided to use the gtk image from Ubuntu.
Why not containerise everything? You need libreoffice? No problem, here is a docker or podman container.
Flatpak is basically GUI-optimized containers. It uses the same technology (namespaces) as docker and podman, just with some extra tools to make GUI-related things work properly. That’s why flatpak apps don’t use the system’s gtk version – they’re running in a sandbox with a different rootfs. You can spawn a shell into the sandbox of a specific app with
flatpak run --command=sh com.yourapp.YourApp
and poke around it if you want to.I’m not too familiar with whatever Android is doing with apks these days tbh. I just don’t like how AppImages fails at the one thing it should do (universality) and doesn’t have the repo model built in. You can have third party solutions to that but it’s just not the same experience.
Why not containerise everything? You need libreoffice? No problem, here is a docker or podman container.
I’ve heard people suggest such a solution. Everything is a container and stuff is just exported out so that it shows up to the system like a normal program. Can’t really say I’m the right person to judge the pros and cons.
There is always another pack installed, freecode, gtk, qt whatever. Even if the system has already the correct gtk version, nope, the dev decided to use the gtk image from Ubuntu.
It can be both good and bad and sometimes it’s necessary. The whole system relies on being able to use different versions of libraries. But having them as separate packs can help in that programs can share those packs so as a dev you can just target one common base and have your stuff work everywhere. And sharing those runtimes has the benefit of someone else keeping it up to date while you can just test if the updated version works for you and switch to that if it does and so on. And with deduplication, runtimes and stuff share the parts that are common to both afaik.
It’s a bit more complicated than just shoving everything in but also it’s less work than same thing having to be packaged separately for every distro.
If you really need that software couldn’t you just use the Windows version?
Unpopular opinion: snap is not so bad and genuinely useful for many things
I would rather have a snap than building from source or use some tar.gz archive with a sketchy install script
I would rather have a snap than building from source or use some tar.gz archive with a sketchy install script
I agree, but that sounds like false dichotomy to me because snap competes with flatpak.
I never presented this as a dichotomy. You know, people prefer things in a certain order, right? I prefer Flatpaks and native packages over snaps and I prefer snaps to building from source.
True, but your post did kinda read like this:
There are plenty of use cases that snap provides that flatpak doesn’t - they only compete in a subset of snap’s functionality. For example, flatpak does not (and is not designed to) provide a way to use it to distribute kernels or system services.
I don’t think that the distribution of system packages is the issue. People need a way to easily distribute and obtain everyday applications, and to keep them up to date in the same manner. Linus spoke about this: https://www.youtube.com/watch?v=Pzl1B7nB9Kc
It depends what you’re trying to accomplish. For me, having the ability to essentially use Lego to put together my system is one of the great features of both snap and nix that Flatpak doesn’t cover.
deleted by creator
some tar.gz archive with a sketchy install script
I just can’t… like maybe I’m too old and that’s why I still can’t wrap my head around how we went from “./configure && make & make install scripts are almost the de facto way to install software in linux” to “a sketchy install script”. We’re living interesting times at Linux
yeah idk a multi thousand line
configure
script seems sketchy to me, like what happened with xzBlame the thousands of supply chain attacks.
Last time I ran a corporate-made installer, it caused massive graphical glitches and lock-ups after waking from sleep. It basically gave my system computer-AIDS.
That’s why I never run scripts which are too long for me to easily understand outside a sandbox. Official distro repositories and Flatpaks are the only sources I have some level of trust in.
I remember those times too. The difference today is that there are so many more libraries and projects use those libraries a lot more often.
So using configure and make means that the user also has the responsibility of ensuring all those libraries are up to date. Which again if we’re talking about not using binary install, each also need a regular configure/make process too. It’s not that unusual for large packages to have dependencies on 100+ libraries. At which point building and maintaining the build for all of them yourself becomes untenable really. However I think gentoo exists to automate a lot of this while still building from source.
I understand why binaries with references to other binary packages for prerequisites are used. I also understand where the limits of this are and why the AppImage/Flatpak/snaps exist. I just don’t particularly like the latter as a concept. But accept there’s times you might need them.
Very unpopular
I would prefer manually writing each software using butterflies over having
snapd
installed on my system.obligatory “there is always a relevant xkcd”
And what do they offer over flatpak?
Nothing useful for me. Given the choice I will usually pick the flatpak.
Better cli experience and the permission prompts are two that come to mind.
A built-in way to have services running (which is why openprinting can make a snap of CUPS but AFAICT can’t make a Flatpak).
I’d rather be able to use my web browser uninterrupted without it being updated while using it and be forced to restart it.
The updates download in the background and will install when you exit the snapped app. If you really don’t want automatic updates, you can run
snap refresh --hold
to hold all automatic updates or add a snap name to hold updates for that snap.Nope. There have been multiple times where I have my browser open, in the middle of something and when I go to open a new tab/window I get a blank screen telling me I need to restart FF to continue.
That is the behaviour that’s built for when an upgrade through a “classic” package manager (e.g. apt, dnf) updates Firefox while it’s still running. The only way I can think of that you’d get that with a snap is if you’re intentionally bypassing the confinement (e.g. by running
/snap/firefox/current/usr/lib/firefox/firefox
directly, which can also massively mess with other things since Firefox won’t be running in thecore22
environment it expects).If you’re using the snap as expected (e.g. opening the
.desktop
file in/var/lib/snapd/desktop/applications/
, running/snap/bin/firefox
or runningsnap run firefox
), snapd won’t replace/snap/firefox/current
until you no longer have any processes from that snap running. Instead you’ll get a desktop notification to close and restart Firefox to update it, and two weeks to either do so or to runsnap refresh --hold firefox
to prevent the update (or something likesnap refresh --hold=6w firefox
to hold the refresh for 6 weeks). Depending on what graphical updater you have, you may also have the ability to hold the update through that updater.Are you sure you’re running the Firefox snap? Because that sounds pretty much precisely like the expected behaviour if someone had gone to lengths to avoid using the snap.